part 2 Flashcards
Azure Cognitive services
Comprehensive family of AI services + cognitive APIs to help build intelligent apps:
- Create customisable pretrained models
-Deploy cognitive services anywhere from cloud to the edge with containers
- No ML expertise needed
- Strict ethical standards
Azure cognitive services: decision
Anomaly detector
Content Moderator
Personaliser - rich personalised experiences for every user
Azure cognitive services: language
Language understanding
QnA Maker
Text analytics - detect sentiment, key phrases, named entities
Translator
Azure cognitive services: speech
speech to text
text to speech
speech translation
speaker recognition
Azure cognitive services: vision
Computer vision
Custom vision
Face
How is cognitive services an umbrella AI service
Enable access multiple AI services w API key and API endpoint
knowledge mining
discipline in AI using combination of intelligent services to quickly learn from vast amounts of information
-allow understanding and exploration of information, find hidden insights, relationships and patterns at scale
process of knowledge mining
ingest - structured: databases, csv. unstructured: pdf, video, image, audio
enrich - enrich with AI capabilities to extract information, find patterns + deepen understanding. cognitive services: vision, language, speech, decision, search.
explore - newly indexed data explore w search, bots, apps, visualisations e.g.CRM, Power BI
Knowledge Mining Use Cases
Content research: helps employees quickly review dense materials
Auditing, risk, compliance management: Help attorneys quickly identify entities of importance from discovery docs + flag important ideas
Business process management: If diagnosis of problem must be in near real-time/fierce bidding competition, avoid costly mistakes
Customer support: find right answer + assess customer sentiment at scale
Contract management: Scour thousands pages of sources to create accurate bid
Azure Face Service
Detect, recognise, analyse human faces:
-detect face (+with special attributes)
-face landmarks
-similar faces
-same face across gallery of images
Face ID - unique ID string for each detected face using 27 predefined landmarks
Face Service Face Attributes
- Accessories
- Age
- Blur
- Emotion
- Facial Hair
- Gender
- Glasses
- Hair
- Head Pose
Azure Translate Service
- 90 languages and dialects
- Neural Machine Translation replacing statistical machine translation
- Custom translator: extend serice for translation based on business/domain use case
Azure Speech service
Speech to text real time
Batch Speech to text
Multi device conversation
Conversation transcription
Custom Speech Models
Text to speech: using speech synthesis markup language, create custom voices
Voice assistance: integrates with bot framework SDK
Speech recognition: speaker verification + identification
Azure Text Analytics API
NLP service for text mining + text analysis:
sentiment analysis, opinion mining, key phrase extraction, language detection, named entity recognition (identify + categorise entities as places, people etc). Subset is Personally identifiable info
Key phrase extraction
Identify main concepts of text quickly.
Works best with bigger amounts of text.
Doc size has max 5120 characters and can have up to 1000 items per collection
Named entity recognition
Detects words and phrases mentioned in unstructured text that can be associated with one or more semantic types
e.g. location, event, location etc
Sentiment analysis
Apply labels and confidence score to text at the sentence and document level: negative, positive, mixed, neutral.
Confidence scores 0-1
Opinion mining = more granular data, Subject and Opinion tied to a sentient
OCR
Extract printed or handwritten text into digital and editable format
e.g. street signs, products, documents etc
Azure has OCR API and Read API - performed by computer vision SDK
OCR API
Older recognition model, only supports images, executes synchronously so suited for less text, support more languages, easier to implement
Read API
Updated recognition model, supports images and PDFs, executes asynchronously, parallelizes tasks per line for faster results, suited for lots of text. Supports fewer languages and harder to implement.
Form recogniser Service
Specialised OCR service. Translates printed text into digital and editable content and preserves that structure and relationship of form-like data.
- Use to automate data entry in apps and enrich document search capabilities
Can identify: - key value pairs, selection marks, table structures
Output structures as: - original file relationships, bounding boxes, confidence scores
Form recogniser composed of
Custom document processing models, prebuilt model for invoices/receipts/ids/business cards, the layout model
